Weekly GitHub Report for Tensorflow: August 25, 2025 - September 01, 2025 (12:01:21)

Weekly GitHub Report for Tensorflow

Thank you for subscribing to our weekly newsletter! Each week, we deliver a comprehensive summary of your GitHub project's latest activity right to your inbox, including an overview of your project's issues, pull requests, contributors, and commit activity.

---

Table of Contents

• I. News
  • 1.1. Recent Version Releases
  • 1.2. Other Noteworthy Updates
• II. Issues
  • 2.1. Top 5 Active Issues
  • 2.2. Top 5 Stale Issues
  • 2.3. Open Issues
  • 2.4. Closed Issues
  • 2.5. Issue Discussion Insights

• III. Pull Requests
  • 3.1. Open Pull Requests
  • 3.2. Closed Pull Requests
  • 3.3. Pull Request Discussion Insights

• IV. Contributors
  • 4.1. Contributors

I. News

1.1 Recent Version Releases:

The current version of this repository is v2.19.0

1.2 Version Information:

Released on March 5, 2025, TensorFlow version 2.19.0 introduces breaking changes to the tf.lite API, including the deprecation of tf.lite.Interpreter in favor of ai_edge_litert.interpreter and changes to certain C++ constants for improved API flexibility. Key updates also include runtime support for the bfloat16 data type in the tfl.Cast operation, alongside the discontinuation of standalone libtensorflow package publishing, while still allowing unpacking from PyPI.

II. Issues

2.1 Top 5 Active Issues:

We consider active issues to be issues that that have been commented on most frequently within the last week. Bot comments are omitted.

1. ValueError: Exception encountered when calling Sequential.call(). Cannot take the length of shape with unknown rank.: This issue concerns a ValueError encountered when calling a TensorFlow Sequential model, specifically related to the inability to determine the length of a shape with an unknown rank. The user reports that despite attempts to explicitly set input shapes in both Sequential and Functional APIs, the error persists, indicating that TensorFlow is unable to infer the input tensor shapes properly during model training.

   • The comments focus on diagnosing the error as stemming from unknown input tensor shapes, with suggestions to explicitly set shapes in the data pipeline. The user struggles with the as_list() error when trying to define shapes, and despite verifying input tensor sizes, the problem remains unresolved, highlighting difficulties in ensuring TensorFlow recognizes input dimensions correctly.
   • Number of comments this week: 6

2. v2.20 : erreur [mutex.cc : 452] RAW: Lock blocking: This issue reports a bug encountered when importing TensorFlow version 2.20.0 on a Mac with an ARM CPU, where the user experiences a mutex lock blocking error that does not occur in version 2.19.0. The problem appears related to an interaction between TensorFlow and PyArrow libraries, with import order affecting whether the process hangs or crashes, suggesting a concurrency or initialization conflict in the underlying C++ runtime or threading model.

   • The comments include attempts to reproduce the issue on similar hardware without success, requests for detailed environment information, and reports from another user confirming the problem with specific import orders involving PyArrow and TensorFlow. Diagnostic steps and a triage matrix were proposed to isolate the cause, including removing tensorflow-io, pinning PyArrow versions, and testing TensorFlow versions, along with suggested workarounds such as process separation or delayed imports to avoid the mutex conflict.
   • Number of comments this week: 4

3. TensorFlow master branch missing CUPTI dependencies: This issue reports a bug in the TensorFlow master branch where compiling with clang-18 against a local CUDA installation results in an undefined symbol error related to missing CUPTI sampler implementation files. The user identifies that the samplerImp.cc and header files are absent from the dependencies, causing the build failure on Ubuntu 24.04 with TensorFlow version 2.21.

   • The comments acknowledge the issue and mention that a pull request fixing the missing dependency and operator overload has been merged, but it is clarified that the fix needs to be applied in the OpenXLA/XLA repository rather than TensorFlow itself, as the problem pertains solely to the XLA codebase.
   • Number of comments this week: 2

4. tf.image.decode_image(img, channels=3) throws ValueError: 'images' contains no shape.: This issue reports a bug where calling tf.image.decode_image(img, channels=3) results in a ValueError stating that the 'images' tensor contains no shape, which occurs during image resizing in a TensorFlow data pipeline. The user has reproduced the bug with TensorFlow Nightly on Ubuntu 25.04 using Python 3.13.3 and provided a minimal code snippet and detailed error traceback to illustrate the problem.

   • The comment acknowledges the issue and apologizes for the delay, informing the user that a community member has submitted a pull request intended to fix the problem; the issue is expected to be resolved once the PR is merged.
   • Number of comments this week: 1

5. tf.function(jit_compile=True) fails with GRU/LSTM (CuDNNRNNV3) on XLA GPU JIT: "No registered 'CudnnRNNV3' OpKernel for XLA_GPU_JIT": This issue reports a bug where using tf.function(jit_compile=True) with tf.keras.layers.GRU or LSTM on a GPU causes XLA compilation to fail due to the lack of a registered CudnnRNNV3 OpKernel for XLA_GPU_JIT devices, despite the same code running fine in eager mode. The problem arises because tf.keras uses fused cuDNN RNN operations that are not supported by XLA GPU JIT, whereas standalone Keras 3 uses pure TensorFlow kernels compatible with XLA, allowing successful compilation.

   • The commenter confirmed the issue on TensorFlow versions 2.19.0 and 2.20.0, explained the root cause as unsupported fused cuDNN ops in XLA GPU JIT, and noted that using standalone Keras 3 avoids the problem by employing XLA-compatible kernels, providing example code and references for clarification.
   • Number of comments this week: 1

2.2 Top 5 Stale Issues:

We consider stale issues to be issues that has had no activity within the last 30 days. The team should work together to get these issues resolved and closed as soon as possible.

1. TF-TRT Warning: Could not find TensorRT: This issue describes a problem where TensorFlow on Ubuntu 22.04 cannot detect TensorRT despite having an NVIDIA RTX 3050 Ti GPU with the appropriate CUDA 12.4 and NVIDIA driver 535 installed. The user reports that the system defaults to an older driver version 550 which is incompatible, and even after reverting to the newer driver and reinstalling TensorFlow, the warning "Could not find TensorRT" persists, hindering their ability to run TensorFlow with TensorRT optimizations.
2. SystemError in tf.ensure_shape and tf.compat.v1.ensure_shape when dtype of shape is tf.uint64 and its value is too large.: This issue reports a bug in TensorFlow where using tf.ensure_shape or tf.compat.v1.ensure_shape with a shape tensor of type tf.uint64 containing very large values close to 2^64 causes a SystemError and OverflowError. Specifically, when such large values are passed in eager execution mode, the functions fail with an error related to the built-in isinstance function, indicating improper handling of large unsigned 64-bit integers in shape validation.
3. Feature Request: Integrate different Digital Signal Processing into tf.signal: This issue is a feature request proposing the integration of advanced digital signal processing (DSP) functionalities, similar to those found in the julius library, into TensorFlow's tf.signal module. The requester highlights the current lack of comprehensive audio data augmentation tools within TensorFlow compared to PyTorch and suggests that adding these capabilities would enhance audio model training by enabling native, efficient preprocessing and augmentation workflows.
4. [DOCS] Missing complex input for Round op: This issue reports a documentation bug in TensorFlow where the Round operation is described as supporting complex tensor inputs, but in practice, attempting to use a complex tensor with this operation results in an error, requiring users to round the real and imaginary parts separately. The user provides a reproducible example on MacOS with TensorFlow 2.15.0 and highlights a discrepancy between the official documentation and the actual behavior, including an error indicating that no suitable device kernel was found for the complex input type.
5. tf.raw_ops.Unbatch aborts with "Check failed: d < dims()": This issue reports a bug in TensorFlow version 2.17 where the operation tf.raw_ops.Unbatch causes the program to abort with a "Check failed: d < dims()" error. The problem occurs on Linux Ubuntu 20.04.3 LTS using Python 3.11.8, and the user has provided a minimal reproducible example demonstrating that the error triggers a core dump during execution.

2.3 Open Issues

This section lists, groups, and then summarizes issues that were created within the last week in the repository.

Issues Opened This Week: 12

Summarized Issues:

• TensorFlow image decoding shape error: Calling tf.image.decode_image(img, channels=3) in TensorFlow 2.20.0 on Ubuntu 25.04 with Python 3.13.3 results in a ValueError because the decoded image tensor lacks a defined shape, causing issues during image resizing. This prevents proper handling of image tensors that do not have explicit shape information.
• issues/99333

• XLA JIT compilation failure with RNN layers on GPU: Using tf.function(jit_compile=True) with tf.keras.layers.GRU or LSTM on a GPU causes XLA JIT compilation to fail due to the absence of a registered CudnnRNNV3 OpKernel for XLA_GPU_JIT devices. Despite this, the same code runs successfully in eager mode, indicating a specific limitation in the JIT compilation path.
• issues/99385

• TensorFlow Lite model quantization and fusion issues: After quantization, the transposed convolution operation in TensorFlow Lite models experiences significantly increased execution time, and concatenation nodes do not fuse activation functions during model conversion. These issues highlight performance degradation and missing optimization support in TFLite conversion workflows.
• issues/99434, issues/99450

• QuantizeAndDequantizeV3 operation crash with non-scalar inputs: Using the QuantizeAndDequantizeV3 operation with non-scalar input_min and input_max tensors and axis=-1 causes a fatal CHECK failure that crashes the Python process instead of raising a catchable error. This bug leads to instability and unexpected termination during quantization operations.
• issues/99458

• TensorFlow import mutex lock error on macOS ARM: Importing TensorFlow modules on macOS ARM CPUs in version 2.20.0 triggers a mutex lock blocking error related to conflicts with PyArrow imports. Downgrading to TensorFlow 2.19.0 temporarily resolves this issue, indicating a regression or compatibility problem in the newer version.
• issues/99464

• CUDA CUPTI dependency missing in TensorFlow 2.21 build: Compiling TensorFlow 2.21 on Ubuntu 24.04 with CUDA 12.9.1 fails due to missing CUPTI dependency files (samplerImp.cc and headers), causing undefined symbol errors in the XLA profiler's CuptiPmSampler implementation. This prevents successful build and profiling support on the specified platform and CUDA version.
• issues/99471

• Incorrect handling of minlength and maxlength in tf.math.bincount: The tf.math.bincount function incorrectly handles minlength and maxlength parameters separately rather than together, resulting in erroneous behavior when maxlength is smaller than minlength. This causes unexpected output sizes and potential logic errors in counting operations.
• issues/99513

• Documentation update needed for tf.nn.pool dilations with SAME padding: The docstring for the dilations parameter in tf.nn.pool() requires correction to specify that dilation rates greater than 1 with "SAME" padding raise a ValueError. Additionally, outdated reference links need updating to reflect current behavior and documentation standards.
• issues/99517

• Inconsistent TFLite quantization results across platforms: A model containing an ADD layer that sums outputs from two fully connected layers produces different quantization scales and inconsistent results when converted to TFLite on Windows versus WSL. This discrepancy occurs despite using similar TensorFlow versions, indicating platform-specific quantization behavior.
• issues/99533

• Feature request for data pre-sorting curriculum using Mahalanobis Distance: A proposal suggests developing a curriculum concept for pre-sorting training data inspired by Mahalanobis Distance to potentially improve training processes in TensorFlow 2.20.0. This feature aims to enhance model training efficiency through better data organization.
• issues/99562

• Redundant favicon link in TensorFlow website templates: The TensorFlow website templates include a redundant favicon link to a non-existent image, which is overshadowed by another valid favicon link. Cleaning up this redundant link would avoid confusion and improve website template quality.
• issues/99575

2.4 Closed Issues

This section lists, groups, and then summarizes issues that were closed within the last week in the repository. This section also links the associated pull requests if applicable.

Issues Closed This Week: 10

Summarized Issues:

• Model saving and loading inconsistencies with @tf.function decorator: Multiple issues report that applying the @tf.function decorator or using XLA compilation on subclassed Keras models causes inconsistent outputs before and after saving and loading the model. These problems indicate serialization and restoration bugs related to graph tracing or XLA-compiled functions in TensorFlow 2.19.0.
• issues/98230, issues/98233

• XNNPACK build and tag version errors in TensorFlow Lite: Several issues describe build failures caused by incorrect or ambiguous XNNPACK tag versions and references, including pointing to wrong commits and mismatched repository tags. These errors prevent successful compilation until the XNNPACK branch is updated to a compatible and synchronized version.
• issues/98662, issues/98772, issues/98967

• Binary incompatibility and DLL load failures on Windows: Issues report runtime errors and import failures on Windows 10 due to binary incompatibility between NumPy versions and DLL load failures of internal TensorFlow modules. These problems are linked to environment setup conflicts and incompatible compiled modules, causing TensorFlow runtime loading to fail.
• issues/98406, issues/99316

• Tensor operation inconsistencies and crashes: Bugs include inconsistent results from tf.math.multiply on CPU vs GPU with complex tensors containing infinity and NaN, and a hard crash caused by QuantizeAndDequantizeV3 operation with invalid tensor shapes. These issues highlight problems with tensor shape validation and numerical consistency across devices.
• issues/98408, issues/99456

• Input shape specification errors in TensorFlow layers: One issue describes confusion and errors when specifying 2-dimensional input shapes in TensorFlow, where invalid syntax or incorrect tuple usage leads to syntax errors or value conversion failures. This points to unclear or strict requirements for defining input shapes excluding batch size.
• issues/99543

2.5 Issue Discussion Insights

This section will analyze the tone and sentiment of discussions within this project's open and closed issues that occurred within the past week. It aims to identify potentially heated exchanges and to maintain a constructive project environment.

Based on our analysis, there are no instances of toxic discussions in the project's open or closed issues from the past week.

---

III. Pull Requests

3.1 Open Pull Requests

This section provides a summary of pull requests that were opened in the repository over the past week. The top three pull requests with the highest number of commits are highlighted as 'key' pull requests. Other pull requests are grouped based on similar characteristics for easier analysis. Up to 25 pull requests are displayed in this section, while any remaining pull requests beyond this limit are omitted for brevity.

Pull Requests Opened This Week: 6

Key Open Pull Requests

1. Fixes tf.image.decode_image(img, channels=3) throws ValueError: 'images' contains no shape.: This pull request addresses a ValueError caused by missing shape information in tf.image.decode_image when used in tf.data pipelines by fixing shape inference, adding comprehensive tests to verify the fix across multiple channel counts and edge cases, and ensuring compatibility with existing behavior and various runtime modes.

• URL: pull/99380

• Merged: No

• Associated Commits: 85f40, 75118, 741fb, ad469, 6d228

2. Add SparseTensor support for math operations : This pull request adds and tests dispatcher support for several unary mathematical operations (asinh, sin, tan, asin, atan, atanh) on TensorFlow's SparseTensor data structure, enabling these functions to operate correctly on sparse inputs.

• URL: pull/99322

• Merged: No

• Associated Commits: 1ffcb, 0b9e3, 22666

3. Fix for cuda12.9.1 compilation: This pull request addresses compilation errors encountered when building TensorFlow with CUDA version 12.9.1 or later by modifying the code to resolve issues related to function overloads and memory alignment.

• URL: pull/99364

• Merged: No

• Associated Commits: c59bd, 7a85e

Other Open Pull Requests

• MLIR TOSA dialect updates: This pull request updates the rounding mode, NaN propagation, and resize mode in the MLIR TOSA dialect from string attributes to enumerations to align with recent LLVM changes. These changes improve consistency and compatibility within the dialect.
• pull/99500

• TensorFlow function parameter fixes: These pull requests fix the handling of parameters in TensorFlow functions, including correcting the minlength and maxlength parameters in tf.math.bincount and updating the docstring for the dilations parameter when the padding option is "SAME". Both address specific issues to improve function correctness and documentation clarity.
• pull/99514, pull/99518

3.2 Closed Pull Requests

This section provides a summary of pull requests that were closed in the repository over the past week. The top three pull requests with the highest number of commits are highlighted as 'key' pull requests. Other pull requests are grouped based on similar characteristics for easier analysis. Up to 25 pull requests are displayed in this section, while any remaining pull requests beyond this limit are omitted for brevity.

Pull Requests Closed This Week: 8

Key Closed Pull Requests

1. Fix for missing dependency and function: This pull request addresses issues related to a missing dependency and adds an operator+ overload along with a dereferencing option to fix compilation errors involving pointer arithmetic in CUDA-related code within the TensorFlow project.

• URL: pull/99371

• Merged: 2025-08-28T10:01:31Z

• Associated Commits: b2540, 05344, b15a9, 32af7

2. Support to compile TensorFlow with CUDA version 12.9.1: This pull request aims to add support for compiling TensorFlow with CUDA version 12.9.1 by addressing the removal of cub APIs, modifying alignment sign mapping to use datatype instead of templates, and implementing overloaded operator+ and pointer dereferencing.

• URL: pull/99353

• Merged: No

• Associated Commits: 194b8, aa9bb, f9d3b

3. [spam] Fix linux cpu badge : This pull request aims to fix issue #99255 by replacing the broken Linux CPU Kokoro build badge in the README.md with a public GitHub Actions badge that reliably displays build status and accessible logs, thereby improving transparency and removing broken links.

• URL: pull/99393

• Merged: No

• Associated Commits: 141ad, 052a1

Other Closed Pull Requests

• Dependency and Build Configuration Updates: This topic covers pull requests that update TensorFlow Lite's CMake configuration to align the XNNPACK dependency with the correct upstream commit, ensuring consistency and successful builds. It also includes a revert of an XLA-related change that caused presubmit check failures, maintaining build stability.
• [pull/99134, pull/99498]

• Deprecation and Code Modernization: These pull requests focus on removing deprecated arguments, flags, and aliases, updating docstrings and notices to reflect current usage, and replacing deprecated decorators and calls with modern equivalents. This ensures the codebase stays up-to-date and reduces technical debt related to outdated APIs.
• [pull/99330]

• Documentation and Contributor Guidelines Improvements: This topic includes pull requests that enhance the clarity, organization, and consistency of the CONTRIBUTING.md file and improve function docstrings by providing clearer guidance, fixing formatting issues, and enhancing the developer experience. These changes make the contribution process more accessible and improve documentation quality.
• [pull/99392, pull/99334]

3.3 Pull Request Discussion Insights

This section will analyze the tone and sentiment of discussions within this project's open and closed pull requests that occurred within the past week. It aims to identify potentially heated exchanges and to maintain a constructive project environment.

Based on our analysis, there are no instances of toxic discussions in the project's open or closed pull requests from the past week.

---

IV. Contributors

4.1 Contributors

Active Contributors:

We consider an active contributor in this project to be any contributor who has made at least 1 commit, opened at least 1 issue, created at least 1 pull request, or made more than 2 comments in the last month.

If there are more than 10 active contributors, the list is truncated to the top 10 based on contribution metrics for better clarity.

Contributor	Commits	Pull Requests	Issues	Comments
tensorflower-gardener	95	0	0	0
mihaimaruseac	0	0	0	57
djxnqp	0	0	31	0
Aaraviitkgp	15	2	0	7
ezhulenev	15	0	0	0
psamanoelton	2	1	0	12
springcrane	0	0	8	6
No author found	13	0	0	0
ILCSFNO	4	3	3	3
Venkat6871	1	0	0	11